Method and apparatus for predicting game revenue using revenue forecasting model

Abstract

A method for predicting game revenue using an artificial intelligence-based forecasting model, includes receiving, for a first game, rank information for each preset period on an application platform and past revenue information for each preset period preceding a first target period. A first input dataset is created by processing the rank information up to the first target period and the past revenue information preceding the first target period. A first predicted revenue amount for the first target period of the game is generated by inputting the first input dataset into a pre-trained revenue forecasting model. A second input dataset is then created by processing the first input dataset, rank information for a second target period subsequent to the first target period, and the first predicted revenue amount. A second predicted revenue amount for the second target period is generated by inputting the second input dataset into the revenue forecasting model.

What is claimed is: 
     
         1 . A method for predicting game revenue using a revenue forecasting model, performed by a computing device, comprising the steps of:
 receiving, for a first game, rank information for each preset period on an application platform and past revenue information for each preset period preceding a first target period;   creating a first input data, by processing the rank information for each preset period up to the first target period and the past revenue information for each preset period preceding the first target period, for the first game;   generating a first predicted revenue amount for the first target period of the first game by inputting the created first input data into a pre-trained artificial intelligence-based revenue forecasting model;   creating a second input data, by processing the first input data, the rank information for a second target period which is subsequent to the first target period, and the generated first predicted revenue amount, for the first game; and   generating a second predicted revenue amount for the second target period of the first game by inputting the created second input data into the revenue forecasting model.   
     
     
         2 . The method of  claim 1 , wherein the revenue forecasting model is trained by:
 creating a training dataset by transforming rank information for each preset period on the application platform and past revenue information for each preset period, for each of a plurality of games, into training data comprising a pair of the rank information and the past revenue information for each preset period.   
     
     
         3 . The method of  claim 1 , wherein the creating the first input data comprises:
 applying, for the first game, an inverse operation into a rank for a first period and applying, for the first game, an inverse operation into a rank for a second period preceding the first period;   comparing the inversed rank for the first period and the inversed rank for the second period to generate a difference value between the inversed ranks for the first period and the second period; and   creating the first input data by configuring difference values between inversed ranks for each preset period up to the first target period of the first game, which are generated by the applying and the comparing.   
     
     
         4 . The method of  claim 3 , wherein the creating the second input data comprises:
 creating the second input data by configuring a difference value between an inversed rank for the first target period and an inversed rank for the second target period of the first game, which is generated by the applying and the comparing.   
     
     
         5 . The method of  claim 1 , wherein the creating the first input data comprises:
 transforming, using a trigonometric function, a temporal point within each preset period up to the first target period, into a transformation value including a sine value and a cosine value; and   creating the first input data including the transformation value so that the transformation value is utilized for weight of the revenue forecasting model.   
     
     
         6 . The method of  claim 5 , wherein the preset period includes a plurality of period types including an hour, a day, a week, or a month, and a scope of the temporal point is varied depending on a period type. 
     
     
         7 . The method of  claim 5 , wherein the creating the second input data comprises:
 transforming, using the trigonometric function, the temporal point within each preset period up to the second target period, into the transformation value including the sine value and the cosine value; and   creating the second input data including the transformation value so that the transformation value is utilized for weight of the revenue forecasting model.   
     
     
         8 . The method of  claim 1 , further comprising:
 creating a third input data, by processing the second input data, the rank information for a third target period which is subsequent to the second target period, and the second predicted revenue amount, for the first game; and   generating a third predicted revenue amount for the third target period by inputting the third input data into the revenue forecasting model.   
     
     
         9 . The method of  claim 1 , further comprising, prior to generating the first predicted revenue amount:
 generating predicted revenue amount for each preset period preceding the first target period, by sequentially inputting the rank information and the past revenue information for each preset period preceding the first target period of the first game into the revenue forecasting model;   identifying at least one anomaly period in which a difference value between actual revenue amount and predicted revenue amount exceeds a threshold difference value, by comparing the actual revenue amount included in the past revenue information with the predicted revenue amount for each preset period up to the first target period, for the first game; and   replacing the actual revenue amount for the at least one anomaly period with the corresponding predicted revenue amount.   
     
     
         10 . The method of  claim 9 , further comprising:
 for a first anomaly period among the at least one anomaly period, in which the actual revenue amount is less than the predicted revenue amount, receiving information on at least one other game released within a pre-determined period before the first anomaly period;   generating a competitor game list of the first game including the information on the at least one other released game; and   storing the competitor game list mapped to the first game.   
     
     
         11 . The method of  claim 9 , further comprising:
 for a second anomaly period among the at least one anomaly period, in which the actual revenue amount is greater than the predicted revenue amount, receiving user information regarding new users who joined the first game during the second anomaly period and event information related to revenue of the first game during the second anomaly period;   calculating at least one of an average age or an average income of the new users based on the user information; and   storing at least one of the average age or the average income mapped with the event information related to the first game.   
     
     
         12 . The method of  claim 11 ,
 wherein the event information includes information on external events that occurred outside the first game and information on in-game events that were conducted within the first game.   
     
     
         13 . A computing device for predicting game revenue using a revenue forecasting model, comprising:
 a processor including at least one core;   a memory; and   a network module,   wherein the processor is configured to:   receive, for a first game, rank information for each preset period on an application platform and past revenue information for each preset period preceding a first target period;   create a first input data, by processing the rank information for each preset period up to the first target period and the past revenue information for each preset period preceding the first target period, for the first game;   generate a first predicted revenue amount for the first target period of the first game by inputting the created first input data into a pre-trained artificial intelligence-based revenue forecasting model;   create a second input data, by processing the first input data, the rank information for a second target period which is subsequent to the first target period, and the generated first predicted revenue amount, for the first game; and   generate a second predicted revenue amount for a second target period of the first game by inputting the created second input data into the revenue forecasting model.   
     
     
         14 . A non-transitory computer readable storage medium including a computer program, wherein the computer program causes a processor of a computer device to perform a method for predicting game revenue using a revenue forecasting model, the method comprising the steps of:
 receiving, for a first game, rank information for each preset period on an application platform and past revenue information for each preset period preceding a first target period;   creating a first input data, by processing the rank information for each preset period up to the first target period and the past revenue information for each preset period preceding the first target period, for the first game;   generating a first predicted revenue amount for the first target period of the first game by inputting the created first input data into a pre-trained artificial intelligence-based revenue forecasting model;   creating a second input data, by processing the first input data, the rank information for a second target period which is subsequent to the first target period, and the generated first predicted revenue amount, for the first game; and   generating a second predicted revenue amount for the second target period of the first game by inputting the created second input data into the revenue forecasting model.